Preliminary graphs for results exploration
grantvagle
2020-07-01
knitr::opts_chunk$set(echo = FALSE)1) How are we studying the ecology of ecosystem services?
Study systems, places, methods
Study system
Places
Study type
ES type (grouped) & Multifunctionality
Notes: Most studies fall in single ES type bin, some 2 or 3 ES types. Max 11 ES bins in single study (see below)
Looking at number of studies (bottom row) with number of service types considered (top row)
## .
## 1 2 3 4 5 6 7 8 9 11
## 129 44 26 12 8 2 1 3 1 1Multifunctionality
Can be observed in above plot
Response variables by ES type
Probably a word cloud or another sankey (sankey would require binning similar response var’s which is probably too much to bother with)
2) Biotic drivers of ecosystem services
Ecosystem service providers
ESPs - single, multiple, interacting
Notes: Not sure exactly how this will be used, or if the quality is usable?
ESPs and ES type
Notes: pest pathogens more ‘across species’, climate regulation more NA thus less often ESPs considered, Pollination more ‘multiple ESP species’
Community structure
A ‘Yes’ means that the box for community structure in the Kremen themes question was checked.
Notes: 79 studies look at comm structure influencing function and got included
Community structure - how many stopped at biodiversity?
will have to ask Caitlin about this one (or she might be doing it already?)
Community structure and ES type
Notes: Pest pathogens, Habitat creation, Climate regulation pretty even (which means more Yes than expected), Pollination may have more No’s than expected
3) Abiotic drivers of ecosystem services
Environmental drivers
SANKEY
Human drivers
SANKEY
Mediation through biotic driver
not sure exactly what this means or if/how we could test it?
4) Temporal and spatial scale
This section contains three subsections. First, there are some basic plots of the overall trends of spatio-temporal scale across all papers. Second, there is a section looking at space-time tradeoffs. Third, there is a section on scale biases for Methods used, ES type, and Driver group.
Time considered
Number of years
## Warning: Factor `clean_answer` contains implicit NA, consider using
## `forcats::fct_explicit_na`
Multi scale
Connectivity considered
Spatial extent
Scale biases
In this section, the dashed line represents what we would expect if subgroups were selected at random from the overall distribution from Yes’s and No’s. Thus, bars larger or smaller than the dashed line show that either time or space is more/less likely to be considered in that group. (I’m still working on how to write this up more specifically in a non-confusing way…so let me know if you have questions!). It’s also useful to keep in mind that these proportions can also be considered in absolute terms, not just with respect to the dashed line. So, for example, even if something is more likely than expected to be multi-scale, it still may be the case that less than half of the studies in that group are multi-scale (which would still point to some sort of absolute gap).
For temporal trends, a study was considered as a ‘Yes’ if it used ‘space for time’ to simplify plotting.
Space-time tradeoffs
Time trends and space
Multi-scale and time
In the plot below, the dashed line represents the overall proportion of multi-scale studies given that the study considered temporal trends (since that was the only way to get the number of years).
## Warning: Factor `YrsData` contains implicit NA, consider using
## `forcats::fct_explicit_na`
Methods scale biases
For these biases plots, a single paper can fit into multiple groups (e.g. a paper that used Observational and Model/data simulation methods). The dashed line accounts for this, so that it still serves as a ‘random expectation’, but it is a little complicated to explain (I’m working on putting together a little explanation with an example though!).
TimeTrends
Number of years
Unlike the above plots which double count papers that are in multiple groups, these plots keep papers single counted because it shows the counts and not proportions. These also keep all intersecting groups separate instead of only looking at the groups individually. These plots are a little messier, but I’m working on a way to better show these patterns (if it seems worth it).
## Warning: Factor `YrsData` contains implicit NA, consider using
## `forcats::fct_explicit_na`
Multi-scale
Spatial extent
Unlike the above plots which double count papers that are in multiple groups, these plots keep papers single counted because it shows the counts and not proportions. These also keep all intersecting groups separate instead of only looking at the groups individually. These plots are a little messier, but I’m working on a way to better show these patterns (if it seems worth it).
ES type scale biases
For these biases plots, a single paper can fit into multiple groups (e.g. a paper that used Observational and Model/data simulation methods). The dashed line accounts for this, so that it still serves as a ‘random expectation’, but it is a little complicated to explain (I’m working on putting together a little explanation with an example though!).
TimeTrends
Number of years
Unlike the above plots which double count papers that are in multiple groups, these plots keep papers single counted because it shows the counts and not proportions. These also keep all intersecting groups separate instead of only looking at the groups individually. These plots are a little messier, but I’m working on a way to better show these patterns (if it seems worth it).
## Warning: Factor `YrsData` contains implicit NA, consider using
## `forcats::fct_explicit_na`
Multi-scale
Spatial extent
Unlike the above plots which double count papers that are in multiple groups, these plots keep papers single counted because it shows the counts and not proportions. These also keep all intersecting groups separate instead of only looking at the groups individually. These plots are a little messier, but I’m working on a way to better show these patterns (if it seems worth it).
Driver scale biases
For these biases plots, a single paper can fit into multiple groups (e.g. a paper that used Observational and Model/data simulation methods). The dashed line accounts for this, so that it still serves as a ‘random expectation’, but it is a little complicated to explain (I’m working on putting together a little explanation with an example though!).
TimeTrends
Number of years
Unlike the above plots which double count papers that are in multiple groups, these plots keep papers single counted because it shows the counts and not proportions. These also keep all intersecting groups separate instead of only looking at the groups individually. These plots are a little messier, but I’m working on a way to better show these patterns (if it seems worth it).
## Warning: Factor `YrsData` contains implicit NA, consider using
## `forcats::fct_explicit_na`
Multi-scale
Spatial extent
Unlike the above plots which double count papers that are in multiple groups, these plots keep papers single counted because it shows the counts and not proportions. These also keep all intersecting groups separate instead of only looking at the groups individually. These plots are a little messier, but I’m working on a way to better show these patterns (if it seems worth it).
5) Analytical methods
Feedbacks
## # A tibble: 8 x 2
## clean_answer count
## <fct> <int>
## 1 Ecosystem function -> service providers 13
## 2 Ecosystem function -> service providers,No feedbacks measured directly 1
## 3 Ecosystem function / service -> abiotic drivers 2
## 4 Ecosystem function / service -> abiotic drivers,Ecosystem function -> s… 1
## 5 Ecosystem services -> ecosystem function 3
## 6 Ecosystem services -> service providers 2
## 7 No feedbacks measured directly 249
## 8 Service providers -> abiotic drivers,Ecosystem services -> service prov… 1Non-linearities/thresholds
